A substrate (SOI substrate) having an SOI (Silicon On Insulator) structure is known as a substrate having a single-crystal Si layer on an insulating layer. A device using this SOI substrate has many advantages that cannot be achieved by ordinary Si substrates. Examples of the advantages are as follows.    (1) The integration degree can be increased because dielectric isolation is easy.    (2) The radiation resistance can be increased.    (3) The operating speed of the device can be increased because the stray capacitance is small.    (4) No well step is necessary.    (5) Latch-up can be prevented.    (6) A complete depletion type field effect transistor can be formed by thin film formation.
Since an SOI structure has the above various advantages, researches have been made on its formation method for several decades.
As a method, an SOI structure is formed by bonding a single-crystal Si substrate to another thermally oxidized single-crystal Si substrate by annealing or an adhesive. In this method, an active layer for forming a device must be uniformly thin. More specifically, a single-crystal Si substrate having a thickness of several hundred micron must be thinned down to the micron order or less.
To thin the substrate, polishing or selective etching can be used.
A single-crystal Si substrate can hardly be uniformly thinned by polishing. Especially, in thinning to the submicron order, the variation range is several ten %. As the wafer size becomes large, this difficulty becomes more pronounced.
Selective etching is effective to uniformly thin the substrate. However, the selectivity ratio is as low as about 102, the surface planarity after etching is poor, and the crystallinity of the SOI layer is unsatisfactory.
The present applicant has disclosed a new SOI technique in Japanese Patent Laid-Open No. 5-021338. In this technique, a first substrate obtained by forming a porous layer on a single-crystal Si substrate and a non-porous single-crystal layer on its surface is bonded to a second substrate via an insulating layer. After this, the bonded substrate stack is separated into two substrates at the porous layer, thereby transferring the non-porous single-crystal layer to the second substrate. This technique is advantageous because the film thickness uniformity of the SOI layer is good, the crystal defect density in the SOI layer can be decreased, the surface planarity of the SOI layer is good, no expensive manufacturing apparatus with special specifications is required, and SOI substrates having about several hundred-Å to 10-μm thick SOI films can be manufactured by a single manufacturing apparatus.
The present applicant has also disclosed, in Japanese Patent Laid-Open No. 7-302889, a technique of bonding first and second substrates, separating the first substrate from the second substrate without breaking the first substrate, smoothing the surface of the separated first substrate, forming a porous layer again, and reusing the substrate. Since the first substrate is not wasted, this technique is advantageous in largely reducing the manufacturing cost and simplifying the manufacturing process.
To separate the bonded substrate stack into two substrates without breaking the first and second substrates, the following methods are available: the two substrates are pulled in opposite directions while applying a force in a direction perpendicular to the bonding interface; a shearing force is applied parallel to the bonding interface (for example, the two substrates are moved in opposite directions in a plane parallel to the bonding interface, or the two substrates are rotated in opposite directions while applying a force in the circumferential direction); pressure is applied in a direction perpendicular to the bonding interface; a wave energy such as an ultrasonic wave is applied to the separation region; a peeling member (e.g., a sharp blade such as a knife) is inserted into the separation region parallel to the bonding interface from the side surface side of the bonded substrate stack; the expansion energy of a substance filling the pores of the porous layer functioning as the separation region is used; the porous layer functioning as the separation region is thermally oxidized from the side surface of the bonded substrate stack to expand the volume of the porous layer and separate the substrates; and the porous layer functioning as the separation region is selectively etched from the side surface of the bonded substrate stack to separate the substrates.
Porous Si was found in 1956 by Uhlir et al. who were studying electropolishing of semiconductors (“Electrolytic Shaping of Germanium and Silicon”, A. Uhlir, Bell System Technical Journal Vol.35, pp. 333–347, March, 1956).
Porous Si can be formed by anodizing an Si substrate in an HF solution.
Unagami et al. studied the dissolution reaction of Si upon anodizing and reported that holes were necessary for anodizing reaction of Si in an HF solution, and the reaction was as follows (“Formation Mechanism of Porous Silicon Layer by anodization in HF Solution” T. Unagami, Journal of the Electrochemical Society, Vol. 127, pp. 476–483, 1980T. Unagami).Si+2HF+(2−n)e+→SiF2+2H++ne−SiF2+2HF→SiF4+H2SiF4+2HF→H2SiF6orSi+4HF+(4−λ)e+→SiF4+4H++λe−SiF4+2HF→H2SiF6where e+ and e− represent a hole and an electron, respectively, and n and λ are the number of holes necessary to dissolve one Si atom. According to them, when n>2 or λ>4, porous Si is formed.
The above fact suggests that p-type Si having holes is converted into porous Si while n-type Si is not converted. The selectivity in this conversion has been reported by Nagano et al. and Imai (“Oxidized Porous Silicon and It's Application” K. Nagano et al., (The transactions of the institute of electronics and communication engineers), The Institute of Electronics, Information and Communication Engineers, Vol. 79, pp 49–54, SSD79-9549, 1979), (“A New Dielectric Isolation Method Using Porous Silicon” K. IMAI, Solid-State Electronics, Vol. 24, pp. 159–164, 1981).
However, it has also been reported that n-type at a high concentration is converted into porous Si (“Complete dielectric isolation by highly selective and self-stopping formation of oxidized porous silicon” R. P. Holmstrom and J. Y. Chi, Applied Physics Letters, Vol. 42, pp. 386–388, 1983). Hence, it is important to select a substrate which can be converted into a porous Si substrate independently of p- or n-type.
To form a porous layer, instead of the above anodizing method, for example, a method of implanting ions into a silicon substrate may also be used.
For example, in the method described in Japanese Patent Laid-Open No. 5-021338, i.e., the method of preparing a substrate (to be referred to as a bonded substrate stack hereinafter) by bonding a first substrate having a non-porous layer such as a single-crystal Si layer on a porous layer to a second substrate via an insulating layer, and separating the bonded substrate stack at the porous layer so as to transfer the non-porous layer formed on the first substrate side to the second substrate, the technique of separating the bonded substrate stack at high reproducibility and high yield is very important.